ETL ELT Where transforms run integration server (SSIS, Talend, Python) inside the warehouse (SQL, dbt) Load raw data? no — only cleaned data yes, into a staging/raw layer Best for on-prem DBs, strict schemas, sensitive data masking cloud warehouses (Snowflake, BigQuery, Synapse) Re-processing re-extract from source re-run SQL on stored raw data Layered (medallion) design Raw / Bronze — data exactly as extracted, plus load metadata (loadedat, sourcefile). Staging / Silver — typed, de-duplicated, validated. Mart / Gold — business-ready facts and dimensions for reporting. Pipeline properties to design for Idempotent — re-running a batch produces the same result (upserts, partition overwrite). Incremental — process only new/changed data using a watermark. Observable — row counts in/out…